New VMware private AI infrastructure rethinks Tanzu, again
Broadcom's VMware finally abandons a long effort to unify Kubernetes with Cloud Foundry as it battles cloud and virtualization rivals on multiple fronts.
Private AI infrastructure updates at VMware Explore clarified the vendor's product lines as it confronts competitive challenges, but they marked the end of longstanding efforts to unify Kubernetes with its application platform.
As of last week, the Tanzu Application Platform (TAP) has been refocused on its original Cloud Foundry underpinnings, as the IT industry zeitgeist has shifted strongly to supporting generative AI. The release of TAP version 10.3 on Aug. 26 also brought in a new VMware Tanzu Data Intelligence platform that adds advanced data access for AI applications, including support for data lake and warehouse engines, streaming pipelines, high-concurrency queries and data caching.
The changes to Tanzu establish a clearer separation between IaaS and PaaS within VMware's product lineup, according to Burt Toma, vice president and general manager of cloud automation at VMware. He compared the realignment to the way cloud providers position services such as Google Kubernetes Engine and Google Cloud Run.
"Tanzu platform is a black box -- it has a container runtime, and as an enterprise, you don't need to know what that is," Toma said in an interview last week. "You just provide code, and we put it into production. If you want to use Kubernetes as Kubernetes, then VCF, which includes VKS, is what you use."
VMware now supports containers, Kubernetes and other cloud-native infrastructure such as service mesh under version 9 of its vCloud Foundation (VCF) IaaS products for private cloud, announced at VMware Explore in 2024 and shipped in June. This new version of VCF includes vSphere Kubernetes Service (VKS), initially introduced in April as a rebranding for VMware Tanzu Kubernetes Grid Service and Tanzu Kubernetes. It also allows for side-by-side management of Kubernetes clusters and containers alongside VMs under a new vSphere Supervisor.
The latest chapter in VMware's long history has seen an aggressive streamlining of its products and a new focus on private cloud under Broadcom.
VCF 9, Canonical deal close ranks against OpenShift
Despite customer backlash against Broadcom following the close of its VMware acquisition and subsequent license changes and pricing hikes, the company's infrastructure software revenue grew 25% to $6.6 billion in its most recent earnings report.
However, other IT infrastructure vendors have been able to take advantage of customer defections from VMware that have also happened in the past year, especially Red Hat, which has made its own appeal to enterprises aiming to bring AI infrastructure in-house to reduce costs and protect data privacy. Although it's still catching up in hypervisor features, Red Hat has also renewed competition with VMware in virtual machines with its OpenShift Virtualization product.
The fact that Kubernetes and containers are now manageable under a single interface with VCF 9 could change that dynamic, according to analysts.
"Red Hat has been very focused on the building blocks of AI infrastructure, too, and we see this is paying off for them, as they have also been able to capture many large VMware customers that had workloads that fit very well into Red Hat OpenShift Virtualization, based on KubeVirt," said Rob Strechay, an analyst at TheCube Research. "But not every workload fits so well."
OpenShift Virtualization requires users to migrate to containerized VMs; there are various tools that can ease this migration and make it less disruptive, but it still represents more of a shift than some enterprises want to make, according to video commentary during Explore by Torsten Volk, an analyst at Enterprise Strategy Group, now part of Omdia.
"[VKS] brings [the] ability to run legacy applications that have been in the corporate portfolio for [decades] side by side with cloud-native applications and containerized applications," Volk said in the video. "That also means that you don't need to containerize everything and modernize everything because, as we all know, that's not what enterprises want to do. They only want to modernize when it actually brings customer value, not just so that virtual machines are now containerized."
Analysts also see a partnership launched last week between VMware and Canonical to integrate the Ubuntu Linux OS and containers within VCF as a competitive move against Red Hat.
"VMware needed a well-known Linux to underpin all [its] container products with something more official," said Gary Chen, an analyst at IDC. "This will give them something to compete against the enterprise-supported RHEL [Red Hat Enterprise Linux] as part of OpenShift. Ubuntu is seen as a more progressive distro, so it's an interesting choice."
Meanwhile, the new Tanzu Data Intelligence platform offers capabilities similar to those in Red Hat's OpenShift AI product, including a data lakehouse partnership with IBM. Tanzu Platform 10.3 introduces inbound and outbound webhooks for AI inference calls, and a recent expanded partnership with Nvidia now includes Blackwell GPU support within VCF for AI inference.
A lot of the Red Hat AI pieces are still separate products, so VMware is being pretty aggressive about enabling all its VCF customers for AI by default.
Gary ChenAnalyst, IDC
Red Hat has highlighted its support for AI inference since acquiring Neural Magic in 2024. This led to the launch of a new Red Hat AI Inference Server and an open source distributed inference project called llm-d in May. Llm-d enables AI inference on more affordable, widely available CPUs instead of GPU hardware.
"Red Hat obviously leans toward open source components and also has some interesting differentiators with Neural Magic," Chen said. "VMware is more proprietary, has a deeper infrastructure stack to do integrations with and is now including [AI infrastructure] in VCF. ... A lot of the Red Hat AI pieces are still separate products, so VMware is being pretty aggressive about enabling all its VCF customers for AI by default."
Broadcom's private AI fixation brings challenges
Technically, VMware's private AI infrastructure doesn't have to run only in on-premises data centers – it can also operate in private instances of public cloud infrastructure from major cloud providers. However, the company's messaging clearly aims to move enterprise customers away from public cloud providers, which have caused enterprise IT "PTSD," according to public comments by Broadcom CEO Hock Tan in 2024.
This messaging might not resonate with many enterprises, particularly developers, according to Volk's video comments.
"The undertone is, 'We are better than public cloud … and there's no reason to go to public cloud,' and that's simply not true," Volk said. "Developers like simple APIs and the latest and greatest APIs, and you want to be able to very easily use [them] when Amazon launches something new for you as a developer, instead of deploying your whole LLM privately and building this whole [internal] capability."
By contrast, Red Hat continues to emphasize hybrid cloud and centralized, flexible management between private and public cloud resources with its OpenShift AI product. Dell, Nutanix and HPE private cloud are also gunning for VMware in on-premises AI infrastructure, according to Strechay.
But other industry watchers predict that most AI workloads will remain in the public cloud – where Microsoft and AWS also offer their own conversion tools and replacements for VMware.
Existing VMware customers could face a migration process to take advantage of new features even if they don't switch vendors, said Keith Townsend, president at The CTO Advisor, a Futurum Group company.
"VCF 9 doesn't just give you pieces of a platform -- it installs the entire private cloud stack by default," he said. "The technology may be unified, but the people and processes often aren't, and that's where adoption friction shows up."
Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.
Dig Deeper on Systems automation and orchestration